Exploring the efficacy of phenomic and genomic selection for yield and fruit quality traits in strawberry.

Phenomic selection is a breeding approach that incorporates phenomic data into statistical models to predict new genotypes. There is still much to learn about the efficacy of phenomic selection compared to genomic selection and best practices for its application in different crops. We utilized multispectral imaging of 1122 strawberry (Fragaria × ananassa Duchesne) clones across four consecutive seasons to compare genomic selection and phenomic selection within and across seasons. Phenomic selection within seasons was more predictive than genomic selection for fruit yield but was less predictive than genomic selection for fruit quality traits. Phenomic models incorporating vegetation indices (VI) were 16% more effective than models with independent spectral bands. Models combining both phenomic and genomic data were most effective for across-season prediction of yield-related traits, with average predictive abilities of 56% for fruit size and 57% for yield. Models with single timepoints were 91% as predictive as models with weekly data across the season, but this was largely influenced by the specific timepoint of data capture. Lastly, we show that the predictive ability of phenomic selection increased significantly with the number of clonal replicates in the training set. Overall, these results suggest that phenomic selection is highly effective in strawberry breeding but is dependent on the trait, timepoint of data capture, and level of clonal replication.